JP2000066697A - Voice recognizer and method therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voice recognizer using plural standard patterns under a noisy environment. SOLUTION: This recognizer is provided with a voice input part 11 for inputting voice, a characteristic parameter extracting part 12 for extracting a parameter to be collated with a standard pattern, from the input voice, a collating part 13 for collating the characteristic parameter extracted from the input voice with the standard pattern, a standard pattern data base storing part 18 with plural stored standard patterns, a power source information gaining part 15 having an electric motor and an engine as a power source, and for acquiring power source information for a hybrid vehicle of which a power source is switched based on engine efficiency, a standard pattern selecting part 16 for selecting the standard pattern from the storing part 18 by the power source under operation, a standard pattern storing part 17 for storing the selected standard pattern, and a recognition result outputting part 14 for informing a speaker such as a driver of a recognized result. The standard pattern suitable to a noise in a cabin of the hybrid vehicle is selected thereby with a simple constitution.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

TECHNICAL FIELD The present invention relates to a vehicle interior,
The present invention relates to an apparatus and method for voice recognition in a room of a hybrid vehicle having two or more power sources.

[0002]

2. Description of the Related Art Conventionally, a speech recognition apparatus using a plurality of standard patterns in a noise environment has been disclosed in Japanese Patent Laid-Open No. 5-165.
No. 492 is known. FIG. 4 shows a conventional system configuration diagram. The voice input unit 41 of the voice to be recognized and the input voice signal are subjected to frequency analysis, and the estimated noise spectrum is subtracted from the analyzed voice spectrum.
A noise processing unit 43 that removes noise, an analysis unit 44 that obtains a feature vector of an input voice based on a voice signal obtained by the noise processing unit, and a standard pattern created using noise-superimposed voice data corresponding to a noise environment. A standard pattern selecting unit 45 for storing the noise information of the residual noise that cannot be removed by the noise processing unit and selecting an optimal standard pattern from a plurality of standard patterns based on the noise information; It comprises a matching unit 46 that matches the standard pattern selected by the standard pattern selecting unit with the feature vector and recognizes the input voice.

[0003]

However, in the above-mentioned conventional speech recognition apparatus, there is a problem that a process for removing noise must be performed in order to select a standard pattern. The present invention solves the above-described conventional technique, and selects a standard pattern using only information on a power source of a hybrid vehicle. Therefore, the standard pattern can be selected by simple processing.

[0004]

In order to solve this problem, the present invention provides a voice input unit for inputting a voice generated by a voice generator, and a parameter for collating the input voice with a standard pattern. A feature parameter extraction unit to be extracted; a matching unit to match the feature parameters extracted from the input voice with the standard parameters; a standard pattern database storage unit storing a plurality of standard patterns; an electric motor and an engine as power sources A power source information obtaining unit that obtains information on a power source of a hybrid vehicle that switches power sources according to the efficiency of an engine; and a standard pattern that selects a standard pattern from a standard pattern database storage unit according to a currently operating power source. A selection unit; a standard pattern storage unit for storing the selected standard pattern; A voice recognition device in the indoor of a hybrid vehicle and a recognition result output unit for informing the.

[0005] This makes it possible to select a standard pattern corresponding to a change in vehicle interior noise such as running noise caused by a change in the power source of the hybrid vehicle, using only the power source of the hybrid vehicle. Furthermore, the present invention relates to a voice recognition method for matching an input voice with a standard pattern prepared in advance and outputting the closest vocabulary as a recognition result, comprising an electric motor and an engine as power sources, and A voice recognition method in a room of a hybrid vehicle, wherein a standard pattern is switched according to a power source currently operating of the hybrid vehicle that switches a power source with efficiency. As a result, the difference in vehicle interior noise generated when the power source of the hybrid vehicle is the electric motor and when the power source is the engine can be reduced by simple processing, and stable recognition performance can be obtained. This voice recognition device can be used for, for example, a car navigation device.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described with reference to FIGS. (First Embodiment) FIG. 1 is a configuration diagram showing a first embodiment of the present invention. A first embodiment will be described with reference to the drawings. A voice uttered from a voice utterer such as a driver is converted from a sound wave into a voice signal which is an electric signal by a voice input unit 11 including a medium conversion unit such as a microphone. This voice signal is converted into a feature parameter such as an LPC cepstrum coefficient in the feature parameter extraction unit 12. In the matching unit 13, DP matching, HMM
The feature parameters extracted by the feature parameter extraction unit 12 are compared with the standard patterns stored in the standard pattern storage unit 17 by using such a method. The recognition result is notified to a speaker such as a driver by a recognition result output unit 14 such as a display or a speaker. The standard pattern storage 17 is selected from at least two or more types of standard patterns stored in the standard pattern database storage 18 by the standard pattern selector 16 based on the information obtained from the power source information acquisition unit 15. The standard pattern is stored. The power source information acquisition unit 15 provides information on whether the state of the power source of the hybrid vehicle is an electric motor or an engine, for example, a flag indicating the power source provided in the control system of the power source. The information is obtained by watching, and the information is sent to the standard pattern selecting unit 16 each time the power source is switched or immediately before the voice recognition is performed. The standard pattern database storage unit 18 stores at least two types of standard patterns, a standard pattern for an electric motor as a power source and a standard pattern for an engine as a power source. For example,
The standard pattern for the electric motor is created using a sound signal obtained by adding noise recorded during running of the electric motor to a sound recorded in a quiet environment, or a sound signal recorded during running of the electric motor. The standard pattern for the engine is created using a sound signal obtained by adding noise recorded during running of the engine to a sound signal recorded in a quiet environment, a sound signal recorded during running of the engine, or the like. Each time the current power source information is sent from the power source information acquisition unit 15, the standard pattern selection unit 16 sends a standard pattern suitable for the current power source from the standard pattern database storage unit 18, specifically, If the power source is an electric motor, the standard pattern for the electric motor described above is selected. If the current power source is an engine, the standard pattern for the engine described above is selected. Store.

[0007] When there are two or more power sources currently operating, the standard pattern for the power source having the highest ratio as the power source is used. (Second Embodiment) FIG. 2 is a flowchart showing a second embodiment of the present invention. Step 21 of determining the start of voice recognition determines whether or not a trigger for starting voice recognition, such as pressing down a button for the user to start voice recognition, is applied. When it is determined in step 21 that speech recognition is to be started, when it is determined that speech recognition is to be started, in step 22 in which the power source is determined, for example, what is the power source provided in the control system of the power source? It is determined whether the current power source is an electric motor or an engine by, for example, checking the flag. If the current power source is determined to be an electric motor in step 32 for determining what the power source is, the process proceeds to step 23 for selecting a standard pattern for running the electric motor. The standard pattern selected here is created using a voice signal obtained by adding noise recorded during running of the electric motor to a voice recorded in a quiet environment, a voice signal recorded during running of the electric motor, or the like. If it is determined in step 22 that the power source is an engine, the current power source is an engine, the process proceeds to step 24 for selecting a standard pattern for running the engine. The standard pattern selected here is created using a voice signal obtained by adding noise recorded during engine running to a voice recorded in a quiet environment, or a voice signal recorded during engine running.

When there are two or more power sources currently in operation, a standard pattern for the power source having the highest ratio as the power source is used. (Third Embodiment) FIG. 3 shows a flow of a voice recognition process according to a third embodiment of the present invention. The step 31 of determining the start of the voice recognition determines whether or not a trigger for starting the voice recognition has been applied, such as the user pressing down a button for starting the voice recognition. When it is determined in step 31 that the speech recognition is to be started, when it is determined that the speech recognition is to be started, a step 32 in which the power source is determined is determined.
Then, it is determined whether the current power source is an electric motor or an engine by, for example, checking a flag of the power source provided in the power source control system.
If the current power source is determined to be an electric motor in step 32 for determining what the power source is, the process proceeds to step 33 for selecting a standard pattern for running the electric motor. The standard pattern selected here is created using a voice signal obtained by adding noise recorded during running of the electric motor to a voice recorded in a quiet environment, a voice signal recorded during running of the electric motor, or the like. If it is determined in step 32 to determine what the power source is, the current power source is an engine, the process proceeds to step 34 for selecting a standard pattern for running the engine. The standard pattern selected here is created using a voice signal obtained by adding noise recorded during engine running to a voice recorded in a quiet environment, or a voice signal recorded during engine running. In step 35 of the voice recognition process, the voice recognition is performed using the standard pattern selected in step 33 or 34 in which the power source is determined in step 32 and the standard pattern is selected in step 33 or 34. The step of outputting the recognition result outputs the recognition result in order to notify the result of the step 35 of the voice recognition processing to a voice speaker such as a driver.

When there are two or more power sources currently operating, a standard pattern for the power source having the highest ratio as the power source is used. The vehicle-mounted speech recognition device according to the present invention can be used, for example, in a car navigation device. Specifically, when the driver speaks “details”, the scale of the screen of the car navigation device becomes a detailed map, and when the driver “destinates”, it becomes a screen on which destination settings can be set. Furthermore, in the above example, when the driver says “details”, the speaker notifies the recognition result to the “detailed map” from the speaker attached to the car navigation device. Can be issued.

[0010]

As described above, according to the present invention, a standard pattern database storage unit for storing standard patterns suitable for two power sources specific to a hybrid vehicle, and a standard pattern switching unit for switching standard patterns are provided. With the hybrid, very simple method and configuration, it is possible to obtain an effect that relatively stable recognition performance can be obtained with respect to a sudden change in noise in the room of the hybrid vehicle.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIG. 2 is a block diagram showing a second embodiment of the present invention.

FIG. 3 is a block diagram showing a second embodiment of the present invention.

FIG. 4 shows a conventional system configuration diagram.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 11 Speech input part 12 Feature parameter extraction part 13 Matching part 14 Recognition result output part 15 Power source information acquisition part 16 Standard pattern selection part 17 Standard pattern storage part 18 Standard pattern database 21 Step to judge the start of speech recognition 22 Power source Step of judging what it is 23 Step of selecting a standard pattern for running the electric motor 24 Step of selecting a standard pattern for running the engine 31 Step of judging the start of voice recognition 32 Step of judging what the power source is 33 Step of selecting a standard pattern for running the electric motor 34 Step of selecting a standard pattern for running the engine 35 Step of voice recognition processing 36 Step of outputting a recognition result 41 Voice input unit 42 A / D conversion unit 43 Noise processing unit 44 Analyzer 45 Standard pattern Down selector 46 matching unit

Claims (4)

[Claims] 1. A voice input unit for inputting a voice generated by a voice generator, a feature parameter extracting unit for extracting a parameter for collating with a standard pattern from the input voice, and a feature parameter extracted from the input voice. And a standard pattern database storage unit that stores a plurality of standard patterns, and a power source for a hybrid vehicle that has an electric motor and an engine as a power source, and switches the power source depending on the efficiency of the engine. Power source information obtaining unit for obtaining power source information; standard pattern selecting unit for selecting a standard pattern from the standard pattern database storage unit according to the currently operating power source; and standard pattern storage unit for storing the selected standard pattern And a recognition result output unit for notifying the voice generator of a recognition result. Speech recognition apparatus in the room of Brides vehicle. 2. A speech recognition method for comparing an input speech with a standard pattern prepared in advance and outputting a closest vocabulary as a recognition result, comprising: an electric motor and an engine as power sources; A voice recognition method in a room of a hybrid vehicle, wherein a standard pattern is selected according to a power source currently operating of the hybrid vehicle, the power source of which is switched. 3. A step of inputting a voice generated by a voice generator, a step of extracting a parameter for matching with a standard pattern from the input voice, and a step of extracting a feature parameter and a standard parameter extracted from the input voice. Collating, storing a plurality of standard patterns in a standard pattern database storage unit, and acquiring information on a power source of a hybrid vehicle having an electric motor and an engine as a power source and switching the power source according to the efficiency of the engine. Selecting a standard pattern from a standard pattern database storage unit according to a currently operating power source; storing the selected standard pattern; and informing the voice generator of a recognition result; Speech recognition method in a room of a hybrid vehicle having a vehicle. 4. The step of selecting the standard pattern,
4. The method according to claim 3, wherein the voice recognition is performed only when the voice recognition is started, and the standard pattern is not switched during the voice recognition processing.
The voice recognition method in a room of a hybrid vehicle according to the above.